CN108507787A - Wind power gear speed increase box fault diagnostic test platform based on multi-feature fusion and method - Google Patents

Abstract

The invention discloses Wind power gear speed increase box fault diagnostic test platform based on multi-feature fusion and methods, solve the problems, such as the single detection method that wind power planetary gear speedup box fault diagnosis is faced, random wind loads special operation condition, interference of changing oil, with the high advantageous effect of fault diagnosis accuracy, scheme is as follows：Test platform includes：The experiment sewing platform base of Wind power gear speed increase box to be measured is set, vibration acceleration sensor, noise signal sensor is arranged in Wind power gear speed increase box side to be measured, Wind power gear speed increase box side to be measured is connect with fluid information detection sensor, each sensor is individually connect with data acquisition module, appraisal procedure can be directed to random wind loads operating mode and extract vibration performance index, extract noise characteristic index, for changing oil, interference problem extracts fluid characteristic index, to establish the vibration noise fluid Fusion Features assessment models based on deep learning and DS evidence theories.

Description

Fault diagnosis test platform and method of wind power gear gearbox based on multi-feature fusion

technical field

The invention relates to the technical field of gearbox detection, and more specifically relates to a multi-feature fusion-based fault diagnosis test platform and method for a speed-up gearbox of a wind power gear.

Background technique

With the large-scale construction of wind farms, wind turbines have entered a period of frequent failures. Most of the wind farms are located in remote areas, the distance between the wind turbines is relatively long, and the wind turbine body is tall, so it is not convenient to carry out equipment inspections, and the cost of shutdown and maintenance is also very high. Therefore, most of the current wind farms are implemented every six months. permanent maintenance. At present, the fault warning and diagnosis methods of wind turbines mainly rely on the SCADA system in the main monitoring room of the wind farm. According to the large amount of data collected, all the wind turbines in the wind farm can be centrally monitored and controlled; however, early faults of wind turbines cannot be accurately diagnosed. A more intelligent and reliable fault diagnosis method.

The intelligent monitoring and health maintenance of wind turbines is an important guarantee for the development of wind power. The planetary gear gearbox is the core component of the wind turbine. Under long-term random wind load conditions, the gears in the components are prone to pitting, cracks, wear, etc. Typical early injury, potentially huge harm. According to statistics, gear failures are the most likely to occur in wind power equipment, accounting for 60%, and the time and economic cost of gearbox maintenance are extremely high. Therefore, carrying out research on early fault diagnosis of wind power planetary gearboxes can improve the accuracy of fault diagnosis, prevent problems before they happen, and save time and maintenance costs. It meets the needs of the safe and healthy development of the wind power energy industry, and has broad market demand and industry prospects.

Vibration detection and oil monitoring are commonly used fault diagnosis methods. However, in terms of vibration detection, since the wind turbine is affected by random wind loads, the input speed of the wind turbine planetary gear gearbox fluctuates, so the non-stationary characteristics of the vibration signal of the wind turbine planetary gearbox are very significant. The vibration signal fault feature extraction method cannot analyze the non-stationary signal well; in terms of oil monitoring, the number of abrasive particles in the oil will change sharply when the wind power planetary gear gearbox replaces the lubricating oil (that is, the oil change Interference problem), so the traditional characteristic index of abrasive particles cannot accurately reflect the actual fault state of the wind power planetary gear gearbox.

On the other hand, the traditional fault diagnosis and detection methods of planetary gear gearboxes for wind turbines are often relatively simple, and the data obtained by only a single detection method cannot accurately reflect the fault status of planetary gear gearboxes, and the detection results are uncertain; Moreover, the fault feature information reflected by a single detection method is not comprehensive, and the fault can only be extracted and analyzed from a single level, and it is impossible to conduct a multi-angle and multi-level comprehensive evaluation of the fault status of the planetary gear box. Therefore, there is a lack of a wind power planetary gearbox fault state detection and evaluation method based on multi-information feature fusion, and there is also a lack of a corresponding wind power planetary gear gearbox fault diagnosis test platform to provide convenient test data support.

To sum up, in view of the single detection method, random wind load special working conditions, oil change interference and other problems faced by the fault diagnosis of wind power planetary gear gearbox, a fault diagnosis test platform for wind power planetary gear gearbox and a comprehensive, accurate, An intelligent multi-information feature fusion detection and evaluation method is of great significance.

Contents of the invention

In order to overcome the deficiencies in the prior art, the present invention provides a wind power gear gearbox fault diagnosis test platform based on multi-feature fusion. The test platform uses three detection methods of vibration, oil, and noise to collect fault information in parallel, which can realize Fault information is analyzed and processed to fully characterize the type and degree of gear damage from these three aspects, which not only greatly improves the accuracy of fault diagnosis, but also fully extracts and analyzes fault features from different angles.

In addition, the present invention provides a wind power planetary gearbox fault detection and evaluation method based on vibration-noise-oil feature fusion. This multi-feature information fusion fault detection and evaluation method can extract vibration characteristic indicators for random wind load conditions. , extract the noise feature index, and extract the oil feature index for the oil change interference problem, thus establishing a vibration-noise-oil feature fusion evaluation model based on deep learning and DS evidence theory, which is conducive to improving the fault diagnosis of wind power planetary gear gearboxes comprehensiveness, intelligence and accuracy.

The specific scheme of the wind power gearbox fault diagnosis test platform based on multi-feature fusion is as follows:

Fault diagnosis test platform for wind power gearbox based on multi-feature fusion, including:

The base of the test bench, the surface of the wind power gear speed up box to be tested can be installed, the output shaft of the wind power gear speed up box to be tested is set with a load, the load is connected to the wind power gear speed up box to be tested through the fixed axis gearbox, and the wind power gear speed up The shaft is connected to the servo motor through the reduction box, the servo motor is installed on the base of the test bench through the servo motor mounting seat, and the servo motor is connected to the PLC controller;

The wind power planetary gear gearbox to be tested is connected to the vibration signal detection module, the oil information detection module and the noise detection module respectively; the vibration signal detection module, the oil information detection module and the noise detection module are respectively connected to the PLC controller.

The test platform extracts and analyzes the fault features of non-stationary vibration signals by using the technical means of sampling vibration signals at equal angles, which can effectively reduce the influence of speed fluctuations on the accuracy of fault feature extraction and analysis of the vibration signal of the wind power gearbox.

Further, the oil information detection module includes a temperature sensor, an online dielectric constant sensor, an online viscosity sensor, an online wear grain monitoring sensor and a CMOS wear grain image sensor. Among them, each sensor is installed on the oil pipe T-type tee interface of the wind power planetary gear gearbox lubrication system through threaded connection in turn, and the detected lubricating oil temperature information, lubricating oil water content information, lubricating oil viscosity information, wear and abrasive particles The particle size information and wear abrasive type information are sent to the data acquisition module.

Further, one end of the speed-up box of the wind power gear to be tested is sequentially connected to the fine filter, the oil pump, the coarse filter, the cooler and the oil information detection module through oil pipes, and the oil pipe is connected to the speed-up box of the wind power gear to be tested the other end of the

Further, the oil information detection sensor is threadedly installed at the interface of the oil pipe at the end side of the speed-up box of the wind power gear to be tested.

Further, the vibration signal detection module includes a pulse signal acquisition device and several vibration acceleration sensors. The pulse signal acquisition device is a photoelectric encoder installed on the input shaft of the wind power gear speed-up box to be tested, and the vibration acceleration sensors are respectively installed on the wind power gear speed-up box. The bearing seat and the box body at both ends of the box, the vibration acceleration sensor and the pulse signal acquisition device transmit the collected equiangular resampling vibration signal to the data acquisition module (such as a data acquisition card).

Further, the PLC controller is connected to the servo driver, and the servo driver controls the operation of the servo motor; the servo motor has a built-in encoder, and the built-in encoder feeds back the operating parameters of the servo motor to the PLC controller, thereby realizing the control of the servo motor speed. Closed-loop control of torque.

The industrial computer is connected with the PLC controller, the PLC controller is connected with the oil pump, directly controls the start and stop of the oil pump, and the PLC controller is connected with the load. The industrial computer sends control parameters to the PLC controller, and the PLC controller gives the control parameters after calculation and sends them to the servo driver, and the servo driver controls the operation of the servo motor; the servo motor has a built-in encoder, and the motor The operating parameters are fed back to the control unit, so as to realize the closed-loop control of the speed and torque of the servo motor.

The industrial computer integrates the servo motor speed regulation software system, the load regulation software system, and the fault diagnosis software system. When testing, all the equipment of the test platform can be started through the industrial computer.

Specifically, the input shaft on the left side of the wind power planetary gearbox to be tested is connected to the output shaft of the front reduction box through the second coupling; the input shaft on the left side of the front reduction box is connected to the output shaft through the first coupling The output shaft of the servo motor is connected; the output shaft on the right side of the wind power planetary gear box to be tested is connected to the input shaft on the left side of the fixed-axis gearbox through a third coupling; the output shaft on the right side of the fixed-axis gearbox The shaft is connected with the load through the fourth coupling, and the mounting seat of the fixed-axis gearbox and the load mounting seat are solid structures with waist-shaped guide holes and tapered positioning pin holes.

Further, the wind power gear speed-up box to be tested is installed on the base of the test bench through the mounting seat of the wind power gear speed-up box, the load is set on the base of the test bench through the load mounting seat, and the fixed-axis gearbox passes through the fixed-axis gearbox. The installation seat is arranged on the base of the test bench, and the reduction box is arranged on the base of the test bench through the gearbox seat.

Further, the mounting seat of the wind power gear speed-up gearbox, the load mounting seat, the fixed-axis gearbox mounting seat and the reduction box seat are detachably connected to the test bench base by bolts. Each mounting seat is a solid structure with a waist-shaped guide hole and a tapered positioning pin hole, and the mounting seat is provided with a groove for fixing the corresponding mechanism.

In addition, the test platform also includes a power module, which is separately connected to the servo motor, vibration acceleration sensor, noise signal sensor, oil information detection sensor and PLC controller for power supply.

Through the collection of random wind load data in the wind field, the simulation modeling and analysis of the wind load under different working conditions is carried out, and the corresponding random wind load spectrum is obtained. Because the research object is the wind power planetary gear speed-up box, and there is only one speed torque signal at the input end of the wind power planetary gear speed-up box, in order to simplify the test process and avoid unnecessary waste of resources, it is only necessary to calculate the obtained random wind load spectrum The speed and torque signal at the input end of the wind power planetary gear gearbox in the wind power generator transmission chain. However, the speed and torque signal at the input end of the wind power planetary gear gearbox has the characteristics of low speed and high torque. This signal is not easy to be directly generated and adjusted by mechanical equipment. It can be achieved by adding a reducer at the front end of the wind power planetary gear gearbox. Reverse "speed up and torque down", convert the low-speed high-torque signal into a high-speed small torque signal for control, and use this signal as the original input control signal of the test platform, that is, the random wind that the drive motor of this test platform needs to simulate. load signal. The random wind load signal control system consists of four parts: industrial computer, PLC controller, servo driver and servo motor. When the system is running, input signal parameters in the industrial computer according to the test requirements, the industrial computer sends control parameters to the PLC controller, and the PLC controller gives the control parameters after calculation, so that the servo driver controls the operation of the servo motor; the servo motor is equipped with an encoder , Feedback the operating parameters of the motor to the control unit, so as to realize the closed-loop control of the rotational speed and torque of the servo motor.

During the construction and installation of the test platform, the main assembly error that affects the measurement accuracy of the platform is the coaxiality error generated during the assembly of the drive motor, wind power planetary gear box, and load motor input and output shafts in the test platform. When the coaxiality error does not meet the matching standard, it will bring serious noise, vibration and flexible shock during the operation of the equipment, which will cause irreversible damage to the parts of the rotating machinery and seriously affect the service life of the whole machine. From the point of view of the test, it will bring interference to the collection of test data, and even affect the accuracy of the data. In order to reduce the coaxiality error when the rotating shafts are mated, in addition to optimizing the structure and design parameters and improving the machining accuracy of each mating surface during the design and processing of the test bench installation base, a modular adjustable device is also added . On the base installation modules of all rotating mechanisms, a positioning structure that can be accurately centered in the horizontal direction is designed, which can effectively improve the positioning and debugging work strength of the equipment during the assembly process, and realize precise positioning and installation. The bolt positioning holes on the mounting seat of the wind power planetary gear speed-up gearbox are processed into waist-shaped guide holes so that they can be guided and fine-tuned in the horizontal direction perpendicular to the transmission shaft during installation, and then the wind power gearbox is connected by the positioning pins designed on the mounting seat. The assembly position of the wind power planetary gear gearbox is accurately positioned, and finally the positioning bolts on the mounting seat are tightened to complete the assembly of the wind power planetary gear gearbox mounting seat on the test bench base.

This wind power planetary gear gearbox test platform integrates a variety of fault feature acquisition schemes, uses three detection methods of vibration, oil, and noise to collect fault information in parallel, and analyzes and processes the fault information in the industrial control fault diagnosis system. From this Three aspects fully characterize the fault state of the planetary gearbox.

Every time the input shaft of the wind power planetary gear gearbox rotates a sampling angle, the photoelectric encoder collects a speed pulse, and the pulse signal is transmitted to the industrial computer; the industrial computer receives the speed pulse signal and sends a collection command to the acceleration vibration signal sensor The acceleration vibration signal of the planetary gear gearbox is collected once; the acceleration vibration signal collected at this time is the equiangular resampling vibration signal of the wind power planetary gear gearbox, and the collected equiangular resampling vibration signal is transmitted to Data acquisition module (connected with industrial computer).

In order to overcome the deficiencies in the prior art, the present invention also provides a wind power planetary gearbox fault state detection and evaluation method based on vibration-noise-oil feature fusion, comprising the following steps:

1) Extraction of vibration characteristic indicators for random wind load conditions;

2) Extraction of noise feature indicators;

3) Oil feature index extraction for oil change interference;

4) Establishment of a vibration-noise-oil feature fusion evaluation model based on deep learning and DS evidence theory;

5) Diagnosis and evaluation of the fault status of the wind power planetary gear gearbox.

The concrete steps of described step 1) are as follows:

1-1) Through the vibration signal detection module of the test platform, relatively stable equiangular resampling vibration signal data is obtained;

1-2) Based on the complete set empirical mode decomposition method, the equiangular resampled vibration signal is decomposed into a series of eigenmode components;

1-3) Screen out the optimal IMF signal according to the kurtosis criterion to achieve the purpose of filtering and denoising;

1-4) Perform Fourier transform on the optimal IMF signal to obtain the fault order characteristic spectrum, and use it as the vibration characteristic index for training and constructing the deep neural network evaluation model for the first time.

The concrete steps of described step 2) are as follows:

2-1) Obtain the noise signal data of the wind power planetary gear gearbox through the noise signal detection module of the test platform;

2-2) Obtain the sound pressure level and octave spectrum of the noise signal based on the acoustic calculation and analysis method, and use it as the noise characteristic index;

The concrete steps of described step 3) are as follows:

3-1) Obtain the oil information database through the oil information detection module of the test platform;

3-2) Calculate the percentage of the number of different types of abrasive grains in the total abrasive grain amount based on the ferrographic analysis method;

3-3) Calculate the percentage of the number of abrasive grains with different particle sizes based on the laser particle size analysis method in the total abrasive grain number;

3-4) The characteristics of the distribution proportion of wear particle types and the proportion characteristics of wear particle size distribution, which are less affected by oil change interference, are used as the oil characteristic indicators.

The concrete steps of described step 4) are as follows:

4-1) Establish a training sample set Φ, as shown in formula (1), wherein Φ _x is the xth training sample, and V _x , N _x , O _x represent various single characteristic indicators of the xth training sample respectively: Vibration characteristic index, noise characteristic index, oil characteristic index;

4-2) Based on the significant advantages of deep learning in image recognition, machine learning, and big data processing and analysis, various single feature indicators in the training sample set are used as input to train and build deep neural networks with various single feature indicators. Network evaluation model; the output of the model is the fault status of the wind power planetary gearbox;

4-3) Introduce the identification framework in the DS evidence theory into the deep neural network evaluation model, and refer to the output of the deep neural network evaluation model to determine the fault state identification framework of the wind power planetary gear box Θ={F ₁ ,F ₂ ,…,F _n }, where F ₁ , F ₂ ,…, F _n represent n kinds of fault states of wind power planetary gear box;

4-4) Based on the advantages of DS evidence theory in the fusion of multi-source feature information, design a multi-feature deep learning-DS evidence theory fusion decision rule. The credit distribution function, as shown in formula (2):

m _i (F ₁ , F ₂ , ..., F _n , Θ) = (p _i q _i1 , p _i q _i2 , ..., p _i q _in , 1-p _i ) (2)

In the formula, m _i represents the evaluation result credibility distribution function of the deep neural network model of the i-th single characteristic index, i=1,2,...,k, and k is the total number of characteristic indexes such as vibration, noise, oil, etc. ; p _i represents the evaluation result accuracy rate of the deep neural network model of the i-th single characteristic index; q _ij represents the credibility of evaluating the sample as the j-th fault state by the deep neural network model of the i-th single characteristic index, j =1,2,...,n;

For any fault state F _j in the fault state identification framework Θ, the multi-feature deep learning-DS evidence theory fusion decision rule can be expressed by formula (3) and formula (4):

In the formula, for the deep neural network model of various single feature indicators, the accuracy of the evaluation result of the training sample set can be used as the p _i value; the q _ij value can be determined statistically according to the evaluation results of the deep neural network model.

The concrete steps of step 5) are as follows:

Continuously collect new vibration, noise, and oil test data through the test platform, and then extract their vibration characteristic indexes, noise characteristic indexes, and oil liquid characteristic indexes to form new samples to be tested, and input them into the previously established deep learning and The vibration-noise-oil characteristic fusion evaluation model of DS evidence theory can output the fault state of the planetary gear box at the moment, so as to realize the comprehensive, accurate and intelligent diagnosis and evaluation of the fault state of the wind power planetary gear box.

Compared with prior art, the beneficial effect of the present invention is:

1) The present invention provides a wind turbine planetary gearbox test platform for multi-means detection and comprehensive extraction of fault characteristics under random wind load conditions. The fault information is collected in parallel by various detection methods, and the fault information is analyzed and processed in the industrial control fault diagnosis system, so as to fully characterize the type and degree of gear damage from these three aspects. This multi-information fusion fault diagnosis method not only greatly improves the accuracy of fault diagnosis, but also fully extracts and analyzes fault features from different angles.

2) The present invention provides an assembly-oriented test bench design that can effectively reduce the coaxiality error between the input and output shafts of the rotary mechanism. In the process of designing and processing the installation base of the experimental bench, the structure and design parameters are optimized, the processing accuracy of each mating surface is improved, and a modular adjustable device is added. On the base installation modules of all rotating mechanisms, a positioning structure that can be accurately centered in the horizontal direction is designed, which can effectively improve the positioning and debugging work strength of the equipment during the assembly process, and realize precise positioning and installation.

3) The present invention provides a vibration detection module of a wind power generator planetary gear speed-up box test platform based on random wind load. The influence of the accuracy of the fault feature extraction and analysis of the vibration signal of the wind power planetary gear gearbox;

4) The present invention provides an oil detection module, including a temperature sensor, an online dielectric constant sensor, an online viscosity sensor, an online abrasive grain monitoring sensor and a CMOS abrasive grain image sensor, which are respectively installed in oil pipes of the gearbox lubrication system through threaded connections On the T-shaped tee interface to realize the online monitoring of the lubrication state and wear state (including wear type and wear degree) of the wind power gear gearbox lubrication system. At the same time, this installation method is beneficial for maintenance and replacement of faulty sensors to realize the detection module upgrade.

5) The method of adding a reducer at the front end of the wind power gear speed increaser box is used for reverse "speed increase and torque reduction", and the low-speed high-torque signal is converted into a high-speed small torque signal for control, which simulates the random wind load well The low-speed and high-torque signal at the input end of the wind power gear speed-up box in the wind power generator transmission chain calculated by the spectrum calculation, and this signal is used as the original input control signal of the test platform. The test process is simplified and unnecessary waste of resources is avoided.

6) The present invention provides a wind power planetary gearbox fault state detection and evaluation method based on vibration-noise-oil feature fusion. This multi-feature information fusion fault detection and evaluation method can extract vibration feature indicators for random wind load conditions , extract the noise feature index, and extract the oil feature index for the oil change interference problem, thus establishing a vibration-noise-oil feature fusion evaluation model based on deep learning and DS evidence theory, which is conducive to improving the fault diagnosis of wind power planetary gear gearboxes comprehensiveness, intelligence and accuracy.

Description of drawings

The accompanying drawings constituting a part of the present application are used to provide further understanding of the present application, and the schematic embodiments and descriptions of the present application are used to explain the present application, and do not constitute improper limitations to the present application.

Fig. 1 is a schematic diagram of the front structure of the present invention;

Fig. 2 is a schematic view of the top view structure of the present invention;

Fig. 3 is a schematic diagram of the lubrication system of the wind power gear gearbox;

Fig. 4 is a schematic diagram of an oil information detection module;

Fig. 5 is a schematic diagram of the data acquisition system;

Fig. 6 is a schematic diagram of the servo motor control system;

Figure 7 is a schematic diagram of the design of the wind power gear gearbox fault diagnosis test platform.

Fig. 8 is a flow chart of fault state detection and evaluation of wind power planetary gearbox based on vibration-noise-oil feature fusion;

Among them, 1. Load fixing frame; 2. Load; 3. The fourth coupling; 4. Fixed shaft gearbox; 5. The third coupling; ; 8, the second shaft coupling; 9, the front reduction box; 10, the first shaft coupling; 11, the servo motor; 12, the servo motor fixing frame; 13, the test bench base; 14, the servo motor mounting seat; 15. Mounting seat of the front reduction gearbox; 16. Mounting seat of the wind power gear speed-up box to be tested; 17. Mounting seat of the fixed-axis gearbox; 18. Load mounting seat; 19. Photoelectric encoder; 20. Photoelectric encoder mounting seat; 21. Fifth coupling; 22. PLC controller; 23. Oil information sensor; 24. Noise signal sensor; 25. Data acquisition module; 26. Industrial computer; 27. Servo drive; 28. Relief valve; 29 , Fine filter; 30, Oil pump; 31, Coarse filter; 32, Cooler; 33, CMOS abrasive particle image sensor; 34, Online abrasive particle monitoring sensor; 35, Online viscosity sensor; 36, Online dielectric constant sensor; 37. Temperature sensor.

Detailed ways

It should be pointed out that the following detailed description is exemplary and intended to provide further explanation to the present application. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

It should be noted that the terminology used here is only for describing specific implementations, and is not intended to limit the exemplary implementations according to the present application. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural, and it should also be understood that when the terms "comprising" and/or "comprising" are used in this specification, they mean There are features, steps, operations, means, components and/or combinations thereof.

As introduced in the background technology, there are deficiencies in the prior art. In order to solve the above technical problems, this application proposes a wind power gear gearbox fault diagnosis test platform based on multi-feature fusion.

In a typical implementation of the present application, as shown in FIG. 1 , the wind power gear gearbox fault diagnosis test platform based on multi-feature fusion includes a test bench base 13 on which a wind power gear booster box is installed. The gearbox mounting base 16; the wind power gear speed-up box mounting base 16 is equipped with the wind power gear speed-up box 6 to be tested, and the input shaft on the left side of the wind power gear speed-up box 6 is connected to the front deceleration box through the second coupling 8. The output shaft of the box 9 is connected; the input shaft on the left side of the front reduction box 9 is connected with the output shaft of the servo motor 11 through the first coupling 10; The shaft joint 5 is connected with the input shaft on the left side of the fixed-axis gearbox 4; the output shaft on the right side of the fixed-axis gearbox 4 is connected with the load through the fourth coupling 3; the input shaft on the right side of the fixed-axis gearbox 4 extends the end of the shaft It is connected with the photoelectric encoder 19 through the fifth coupling 21 .

The power supply module provides power for the entire device of the wind power gear gearbox fault diagnosis test bench; the load 2 selects a magnetic powder brake.

Servo motor mounting seat 14, front reduction gearbox mounting seat 15, wind power gear speed increasing gearbox mounting seat 16, fixed axis gearbox mounting seat 17 and load mounting seat 18 are solid structures with waist-shaped guide holes and tapered positioning pin holes .

The servo motor 11 is fixed on the servo motor fixing frame 12 through bolt connection, and the servo motor fixing frame 12 is fixed on the servo motor mounting base 14 through bolt connection.

The front reduction box 9 is fixed on the front reduction box installation seat 15 by means of bolt connection. The wind power gear speed-up box 6 is fixed on the wind power gear speed-up box installation seat 16 by means of bolt connection. The fixed-axis gearbox 4 is fixed on the fixed-axis gearbox mounting seat 17 by means of bolt connection. The photoelectric encoder 19 is fixed on the photoelectric encoder mounting base 20 by means of bolt connection. The load 2 is fixed to the load fixing frame 1 by means of bolt connection, and the load fixing frame 1 is fixed to the load mounting seat 18 by means of bolt connection.

The servo motor mounting base 14, the front reduction box mounting base 15, the wind power gear speed increasing box mounting base 16, the fixed axis gear box mounting base 17, the photoelectric encoder mounting base 20 and the load mounting base 18 are all fixed on the Test bench base 13.

One end of the speed-up box 6 of the wind power gear to be tested is connected to the overflow valve 28, the fine filter 29, the oil pump 30, the coarse filter 31, the cooler 32 and the oil information detection module in turn through the oil pipe, and finally the oil pipe is connected to the wind power to be tested. The other end of the gear speed-up box 6 constitutes a wind power gear speed-up box lubrication system; the oil pump 30 is connected with the Advantech IPC-610L industrial computer 26 .

The noise signal detection module includes a noise signal sensor 24. The noise signal sensor 24 is installed in the box of the speed-up box of the wind power gear to be tested, and transmits the collected noise signal to the data acquisition module 25, and then to the Advantech IPC-610L industrial computer 26 .

The oil information detection module includes a temperature sensor 37, an online dielectric constant sensor 36, an online viscosity sensor 35, an online abrasive grain monitoring sensor 34 and a CMOS abrasive grain image sensor 33, which are installed in the wind power gear speed-up box through screw connections in sequence On the T-type tee interface of the oil pipe of the lubrication system, and transmit the detected lubricating oil temperature information, lubricating oil moisture content information, lubricating oil viscosity information, wear abrasive grain size information and wear abrasive grain type information to Siemens S7-200PLC The data acquisition module 25 of the controller 22 is sent to the Advantech IPC-610L industrial computer 26.

Described vibration signal detection module comprises pulse signal acquisition device, several vibration acceleration sensors 7, and pulse signal acquisition device is installed on the photoelectric encoder 19 on the input shaft of wind power planetary gear speed-up box to be measured, and vibration acceleration sensor 7 is respectively installed on On the bearing seats and the box body at both ends of the wind power planetary gear speed-up box 6 to be tested. The vibration acceleration sensor and the pulse signal acquisition device transmit the collected equiangularly resampled vibration signals to the data acquisition module 25 , and then to the Advantech IPC-610L industrial computer 26 .

Advantech IPC-610L industrial computer sends control parameters to the Siemens S7-200PLC controller, and the Siemens S7-200PLC controller gives the control parameters after calculation and sends them to the servo driver 27, which controls the operation of the servo motor 11; the servo motor 11 has a built-in encoder, which feeds back the operating parameters of the motor to the control unit, thereby realizing closed-loop control of the rotational speed and torque of the servo motor 11.

When testing the speed-up box 6 of the wind power gear to be tested, first design a suitable mounting seat 16 of the speed-up box for the wind power gear to be tested according to the size of the speed-up box 6 of the wind power gear to be tested, and adjust the mounting seat 16 of the speed-up box for the wind power gear to be tested and the servo The positions of the motor mount 14, the front reducer mount 15, the fixed-axis gearbox mount 17 and the load mount 18 on the test bench base 13 are respectively adjusted and fixed on the test bench base 13 along the waist-shaped guide hole, Then the wind power gear speed-up box 6 to be measured is placed on the wind power gear speed-up box installation seat 16, fixed with bolts, and the input shaft and the output shaft of the wind power gear speed-up box 6 to be measured are respectively connected with the second shaft coupling 8 Connect with the third coupling 5, and finally connect the lubricating oil inlet/outlet of the wind power gear speed-up box 6 to be tested with the corresponding oil pipe in the wind power gear speed-up box lubrication system, and place the vibration acceleration sensor 7 on the wind power speed-up box to be tested. The corresponding position of the gear speed-up box 6. The power module provides power for the entire device of the wind power gear speed increaser gearbox fault diagnosis test bench.

Advantech IPC-610L industrial computer 26 integrates servo motor speed regulation software system, load regulation software system, and fault diagnosis software system. When testing, all the equipment of the test platform can be started by Advantech IPC-610L industrial computer.

Simultaneously, photoelectric encoder 19, noise signal sensor 24, temperature sensor 37, on-line dielectric constant sensor 36, on-line viscosity sensor 35, on-line abrasive grain monitoring sensor 34 and CMOS abrasive grain image sensor 33 and vibration acceleration sensor 7 will measure The data is transmitted to the data acquisition module 25 through the data line, and then transmitted to the Advantech IPC-610L industrial computer 26. The Advantech IPC-610L industrial computer 26 will process and analyze the collected data. The information and noise signals are used to detect the speed-up box 6 of the wind power gear. In the present invention, the acceleration vibration signal, the oil information and the noise signal of the speed-up box of the wind power gear are considered comprehensively, and more accurate detection of the speed-up box of the wind power gear is realized.

In addition, the present invention also provides a wind power planetary gearbox fault state detection and evaluation method based on vibration-noise-oil feature fusion, including the following steps:

1) Extraction of vibration characteristic indicators for random wind load conditions;

2) Extraction of noise feature indicators;

3) Oil feature index extraction for oil change interference;

4) Establishment of a vibration-noise-oil feature fusion evaluation model based on deep learning and DS evidence theory;

5) Diagnosis and evaluation of the fault status of the wind power planetary gear gearbox.

Specific steps are as follows:

1) Extraction of vibration characteristic indicators for random wind load conditions;

Firstly, through the vibration signal detection module of the test platform, the relatively stable equiangular resampling vibration signal data is obtained; then, based on the complete ensemble empirical mode decomposition method (complete ensemble empirical mode decomposition with adaptive noise, referred to as CEEMDAN), the equiangular resampled The sampled vibration signal is decomposed into a series of intrinsic mode functions (IMF for short); then, the optimal IMF signal is screened out according to the kurtosis criterion to achieve the purpose of filtering and denoising; finally, the optimal IMF signal is calculated by Fourier The characteristic spectrum of the fault order is obtained by Liye transformation, and it is used as the vibration characteristic index for training and constructing the deep neural network evaluation model for the first time.

2) Extraction of noise feature indicators;

First, through the noise signal detection module of the test platform, the noise signal data of the wind power planetary gear gearbox is obtained; then, based on the acoustic calculation and analysis method, the sound pressure level and octave frequency spectrum of the noise signal are obtained, and used as the noise characteristic index.

3) Oil feature index extraction for oil change interference;

When the lubricating oil is changed for the wind power planetary gear speed-up box, the number of abrasive particles in the oil will change sharply (that is, the problem of oil change interference), so that the traditional characteristic index of the number of wear particles cannot accurately reflect the actual fault state of the wind power planetary gear speed-up box ; but in the same wear state, abrasive particles of different wear types and particle sizes will still enter the oil according to the proportion before oil change, so the characteristic index of the abrasive particle type ratio and abrasive particle size ratio in the extracted oil is The breakthrough point to solve the problem of oil change interference.

First, through the oil information detection module of the test platform, the oil information database is obtained; then, based on the ferrography analysis method, different types of abrasive particles (normal wear abrasive particles, severe sliding abrasive particles, cutting abrasive particles, and fatigue wear abrasive particles) are calculated. , oxidized abrasive grains) accounted for the percentage of the total abrasive grains; then, based on the laser particle size analysis method, the abrasive grains of different particle sizes (0-10μm, 10-30μm, 30-50μm, 30-50μm, 50-100μm, particle size above 100μm) accounted for the percentage of the total number of abrasive particles; finally, the proportion characteristics of the type of wear particles and the proportion characteristics of the distribution of wear particle sizes that are less affected by oil change interference are used as oil characteristic indicators .

4) Establishment of a vibration-noise-oil feature fusion evaluation model based on deep learning and DS evidence theory;

In order to improve the utilization rate of the above-mentioned vibration, noise, and oil characteristic indexes, the sensitivity advantages of vibration characteristic indexes in fault location analysis, the sensitivity advantages of noise characteristic indexes in noise source location and the advantages of oil characteristic indexes in fault quantitative analysis are integrated. Sensitive advantages, more comprehensive, accurate, and intelligent evaluation of the fault status of planetary gear gearboxes require the comprehensive application of intelligent technologies such as machine learning and multi-source information fusion to establish an effective vibration-noise-oil feature fusion evaluation model. The specific plan is as follows:

(1) Establish a training sample set Φ, as shown in formula (1), wherein Φ _x is the xth training sample, V _x , N _x , O _x represent various single characteristic indicators of the xth training sample: vibration Characteristic index, noise characteristic index, oil characteristic index.

(2) Based on the significant advantages of deep learning in image recognition, machine learning, and big data processing and analysis, various single feature indicators (vibration feature indicators, noise feature indicators, and oil feature indicators) in the training sample set are used as input Quantity, train and build a deep neural network evaluation model of various single feature indicators; the output of the model is the fault status of the wind power planetary gear box (such as gear wear, gear cracks, bearing wear, etc.).

(3) Introduce the identification framework in the DS evidence theory into the deep neural network evaluation model, and refer to the output of the deep neural network evaluation model to determine the fault state identification framework of the wind power planetary gear box Θ={F ₁ ,F ₂ ,…,F _n }, where F ₁ , F ₂ ,…, F _n represent n kinds of fault states of wind power planetary gearbox.

(4) Based on the advantages of DS evidence theory in the fusion of multi-source feature information, design a multi-feature deep learning-DS evidence theory fusion decision rule. Degree distribution function, as shown in formula (2):

m _i (F ₁ , F ₂ , ..., F _n , Θ) = (p _i q _i1 , p _i q _i2 , ..., p _i q _in , 1-p _i ) (2)

In the formula, m _i represents the evaluation result credibility distribution function of the deep neural network model of the i-th single characteristic index, i=1,2,...,k, and k is the total number of characteristic indexes such as vibration, noise, oil, etc. ; p _i represents the evaluation result accuracy rate of the deep neural network model of the i-th single characteristic index; q _ij represents the credibility of evaluating the sample as the j-th fault state by the deep neural network model of the i-th single characteristic index, j =1,2,...,n.

For any fault state F _j in the fault state identification framework Θ, the multi-feature deep learning-DS evidence theory fusion decision rule can be expressed by formula (3) and formula (4):

In the formula, for the deep neural network model of various single feature indicators, the accuracy of the evaluation result of the training sample set can be used as the p _i value; the q _ij value can be determined statistically according to the evaluation results of the deep neural network model.

(5) Through the test sample set, test and correct the deep neural network evaluation model of various single feature indicators, and improve the multi-feature deep learning-DS evidence theory fusion decision rules; thus establish a deep learning and DS evidence theory The vibration-noise-oil feature fusion evaluation model.

5) Diagnosis and evaluation of the fault status of the wind power planetary gear gearbox

Continuously collect new vibration, noise, and oil test data through the test platform, and then extract their vibration characteristic indexes, noise characteristic indexes, and oil liquid characteristic indexes to form new samples to be tested, and input them into the previously established deep learning and The vibration-noise-oil feature fusion evaluation model of DS evidence theory can intelligently and autonomously output the fault status of the planetary gear gearbox at the moment, so as to realize the comprehensive, accurate and intelligent diagnosis of the fault status of the wind power planetary gear gearbox Evaluate.

The above descriptions are only preferred embodiments of the present application, and are not intended to limit the present application. For those skilled in the art, there may be various modifications and changes in the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application shall be included within the protection scope of this application.

Claims (10)

1. The wind power gear gearbox fault diagnosis test platform based on multi-feature fusion is characterized in that it includes: The base of the test bench, the surface of the wind power gear speed up box to be tested can be installed, the output shaft of the wind power gear speed up box to be tested is set with a load, the load is connected to the wind power gear speed up box to be tested through the fixed axis gearbox, and the wind power gear speed up to be tested The input shaft of the box is connected to the servo motor through the reduction box, the servo motor is installed on the base of the test bench through the servo motor mounting seat, and the servo motor is connected to the PLC controller; The wind power planetary gear gearbox to be tested is connected to the vibration signal detection module, the oil information detection module and the noise detection module respectively; the vibration signal detection module, the oil information detection module and the noise detection module are respectively connected to the PLC controller. 2. The wind power gear gearbox fault diagnosis test platform based on multi-feature fusion according to claim 1, wherein the oil information detection module includes a temperature sensor, an online dielectric constant sensor, an online viscosity sensor, an online Wear particle monitoring sensor and CMOS wear particle image sensor. 3. The wind power gear speed-up box fault diagnosis test platform based on multi-feature fusion according to claim 1, wherein one end of the wind power gear speed-up box to be tested is sequentially connected with a fine filter, an oil pump, and a coarse filter through an oil pipe. The cooler, the cooler, and the oil information detection module are connected, and the oil pipe is connected to the other end of the speed-up box of the wind power gear to be tested. 4. The wind power gear speed-up box fault diagnosis test platform based on multi-feature fusion according to claim 1, wherein the wind power gear speed-up box to be tested is arranged on the test platform through a wind power gear speed-up box mounting seat. The load is installed on the test bench base through the load mounting seat, the fixed-axis gearbox is installed on the test bench base through the fixed-axis gearbox mounting seat, and the reduction box is installed on the test bench base through the reduction box seat. 5. the wind power gear gearbox fault diagnosis test platform based on multi-feature fusion according to claim 1, is characterized in that, described vibration signal detection module comprises pulse signal acquisition device, some vibration acceleration sensors, pulse signal acquisition device For the photoelectric encoder installed on the input shaft of the wind power gear speed-up box to be tested, the vibration acceleration sensors are respectively installed on the bearing seat and the box body at both ends of the wind power gear speed-up box, and the vibration acceleration sensor and the pulse signal acquisition device collect the collected The vibration signal is resampled by angle and sent to the data acquisition module. 6. the wind power gear speed-up box fault diagnosis test platform based on multi-feature fusion according to claim 1, is characterized in that, described PLC controller is connected with servo driver, and servo driver controls described servo motor operation; The motor has a built-in encoder, and the built-in encoder feeds back the operating parameters of the servo motor to the PLC controller, thereby realizing closed-loop control of the speed and torque of the servo motor. 7. A wind power planetary gearbox fault state detection and evaluation method based on vibration-noise-oil feature fusion, characterized in that it comprises the following steps: 1) Extraction of vibration characteristic indicators for random wind load conditions; 2) Extraction of noise feature indicators; 3) Oil feature index extraction for oil change interference; 4) Establishment of a vibration-noise-oil feature fusion evaluation model based on deep learning and DS evidence theory; 5) Diagnosis and evaluation of the fault status of the wind power planetary gear gearbox. 8. A kind of wind power planetary gearbox fault state detection and evaluation method based on vibration-noise-oil characteristic fusion according to claim 7, it is characterized in that, the specific steps of described step 1) are as follows: 1-1) Through the vibration signal detection module of the test platform, relatively stable equiangular resampling vibration signal data is obtained; 1-2) Based on the complete set empirical mode decomposition method, the equiangular resampled vibration signal is decomposed into a series of eigenmode components; 1-3) Screen out the optimal IMF signal according to the kurtosis criterion to achieve the purpose of filtering and denoising; 1-4) Perform Fourier transform on the optimal IMF signal to obtain the fault order characteristic spectrum, and use it as the vibration characteristic index for training and constructing the deep neural network evaluation model for the first time. 9. A kind of wind power planetary gearbox failure state detection and evaluation method based on vibration-noise-oil characteristic fusion according to claim 7, it is characterized in that, the specific steps of described step 2) are as follows: 2-1) Obtain the noise signal data of the wind power planetary gear gearbox through the noise signal detection module of the test platform; 2-2) Obtain the sound pressure level and octave spectrum of the noise signal based on the acoustic calculation and analysis method, and use it as the noise characteristic index; The concrete steps of described step 3) are as follows: 3-1) Obtain the oil information database through the oil information detection module of the test platform; 3-2) Calculate the percentage of the number of different types of abrasive grains in the total abrasive grain amount based on the ferrographic analysis method; 3-3) Calculate the percentage of the number of abrasive grains with different particle sizes based on the laser particle size analysis method in the total abrasive grain number; 3-4) The characteristics of the distribution proportion of wear particle types and the proportion characteristics of wear particle size distribution, which are less affected by oil change interference, are used as the oil characteristic indicators. 10. A kind of wind power planetary gearbox fault state detection and evaluation method based on vibration-noise-oil characteristic fusion according to claim 7, it is characterized in that, the specific steps of described step 4) are as follows: 4-1) Establish a training sample set Φ, as shown in formula (1), wherein Φ _x is the xth training sample, and V _x , N _x , O _x represent various single characteristic indicators of the xth training sample respectively: Vibration characteristic index, noise characteristic index, oil characteristic index; 4-2) Based on the significant advantages of deep learning in image recognition, machine learning, big data processing and analysis, etc., each single feature index in the training sample set is used as input to train and build a deep neural network with various single feature indexes. Network evaluation model; the output of the model is the fault status of the wind power planetary gearbox; 4-3) Introduce the identification framework in the DS evidence theory into the deep neural network evaluation model, and refer to the output of the deep neural network evaluation model to determine the fault state identification framework of the wind power planetary gear box Θ={F ₁ ,F ₂ ,…,F _n }, where F ₁ , F ₂ ,…, F _n represent n kinds of fault states of wind power planetary gear box; 4-4) Based on the advantages of DS evidence theory in the fusion of multi-source feature information, design a multi-feature deep learning-DS evidence theory fusion decision rule. The credit distribution function, as shown in formula (2): m _i (F ₁ , F ₂ , ..., F _n , Θ) = (p _i q _i1 , p _i q _i2 , ..., p _i q _in , 1-p _i ) (2) In the formula, m _i represents the evaluation result credibility distribution function of the deep neural network model of the i-th single characteristic index, i=1,2,...,k, and k is the total number of characteristic indexes such as vibration, noise, oil, etc. ; p _i represents the evaluation result accuracy rate of the deep neural network model of the i-th single characteristic index; q _ij represents the credibility of evaluating the sample as the j-th fault state by the deep neural network model of the i-th single characteristic index, j =1,2,...,n; For any fault state F _j in the fault state identification framework Θ, the multi-feature deep learning-DS evidence theory fusion decision rule can be expressed by formula (3) and formula (4): In the formula, for the deep neural network model of various single feature indicators, the accuracy of the evaluation result of the training sample set can be used as the p _i value; the q _ij value can be determined statistically according to the evaluation results of the deep neural network model.